[Patch] Latest device-mapper snapshot

[Patch] Latest device-mapper snapshot

[Joe Thornber]

New patchballs are available here:

http://people.sistina.com/~thornber/patches/2.5-stable/

Including a diff against 2.5.44-ac1.  There are a lot of changes in
here compared to the last release, however most of these are due to
code refactoring rather than bug fixes.  Highlights include:

o) Make the changes recommended by Christoph Hellwig and others:
   http://marc.theaimsgroup.com/?l=linux-kernel&m=103462345119681&w=2

o) Add reference count to struct mapped_device, and struct dm_table.

o) Hide the above two structs in their respective .c file

o) Move all locking of struct mapped_device into dm.c (we can do this now because
   of the reference counting).

o) Remove the name and uuid field from struct mapped device, these are really
   only used by the interface as a way of refering to devices.

o) Nobody needs to lookup from kdev_t -> struct mapped_device, so remove
   that hash table (thanks to Al Viros recent bdev->bd_disk stuff).

o) dm.c has no need of the dm-hash.c file any more, so merge dm-hash.c into
   dm-ioctl.c (the fs interface uses the dcache for lookups).


There are still open issues that prevent things working perfectly:

o) The gendisk hash table is getting confused when removing a device.  eg, if
   I create 3 devices with minors (1, 2, 3).  Then remove minor 2, get_gendisk 
   will remove minor == 3. (Or I've done something really stupid).

o) Splitting pages still doesn't work, this is a generic block layer
   thing rather than dm.  In practise I can only trigger this with
   striped targets.  So stick to linear targets for now.


Filesystem interface to follow before the end of the week.

- Joe

Re: [linux-lvm] [Patch] Latest device-mapper snapshot

[Austin Gonyou]

YAY!!! :) I will try asap. :-D I will also try with qla2200's and see if
I can break it! :)

On Wed, 2002-10-23 at 05:25, Joe Thornber wrote:
> New patchballs are available here:
> 
> http://people.sistina.com/~thornber/patches/2.5-stable/
> 
> Including a diff against 2.5.44-ac1.  There are a lot of changes in
> here compared to the last release, however most of these are due to
> code refactoring rather than bug fixes.  Highlights include:
> 
> o) Make the changes recommended by Christoph Hellwig and others:
>    http://marc.theaimsgroup.com/?l=linux-kernel&m=103462345119681&w=2
> 
> o) Add reference count to struct mapped_device, and struct dm_table.
> 
> o) Hide the above two structs in their respective .c file
> 
> o) Move all locking of struct mapped_device into dm.c (we can do this
> now because
>    of the reference counting).
> 
> o) Remove the name and uuid field from struct mapped device, these are
> really
>    only used by the interface as a way of refering to devices.
> 
> o) Nobody needs to lookup from kdev_t -> struct mapped_device, so remove
>    that hash table (thanks to Al Viros recent bdev->bd_disk stuff).
> 
> o) dm.c has no need of the dm-hash.c file any more, so merge dm-hash.c
> into
>    dm-ioctl.c (the fs interface uses the dcache for lookups).
> 
> 
> There are still open issues that prevent things working perfectly:
> 
> o) The gendisk hash table is getting confused when removing a device.
> eg, if
>    I create 3 devices with minors (1, 2, 3).  Then remove minor 2,
> get_gendisk 
>    will remove minor == 3. (Or I've done something really stupid).
> 
> o) Splitting pages still doesn't work, this is a generic block layer
>    thing rather than dm.  In practise I can only trigger this with
>    striped targets.  So stick to linear targets for now.
> 
> 
> Filesystem interface to follow before the end of the week.
> 
> - Joe
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

Re: [Patch] Latest device-mapper snapshot

[Joe Thornber]

Doh !

# [Device-mapper]
# Forgot a bio_put()
# --------------------------------------------
#
diff -Nru a/drivers/md/dm.c b/drivers/md/dm.c
--- a/drivers/md/dm.c	Fri Oct 25 10:47:49 2002
+++ b/drivers/md/dm.c	Fri Oct 25 10:47:49 2002
@@ -270,8 +270,10 @@
 	 * finished.  If a partial io errors I'm assuming it won't
 	 * be requeued.  FIXME: check this.
 	 */
-	if (error || !bio->bi_size)
+	if (error || !bio->bi_size) {
 		dec_pending(io, error);
+		bio_put(bio);
+	}
 
 	return 0;
 }